Dual-level Adaptive Self-Labeling for Novel Class Discovery in Point Cloud Segmentation

Ruijie Xu, Chuyu Zhang, Hui Ren, Xuming He

Primary Area: unsupervised, self-supervised, semi-supervised, and supervised representation learning

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Keywords: novel class discovery, point clouds semantic segmantation

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.

Abstract:

We tackle the novel class discovery in point cloud segmentation, which discovers novel classes based on existing knowledge. Existing works propose an online point-wise clustering method with a simplified equal class-size
constraint on the novel classes to avoid degenerate solutions. However, the inherent imbalanced distribution of novel classes
in point clouds contradicts the equal class-size constraint, and point-wise clustering tends to ignore the rich spatial context information of objects, which can result in less expressive representation for semantic segmentation.
To solve the above challenges, we propose a novel self-labeling strategy that adaptively generates high-quality pseudo-labels for imbalanced classes during model training. In addition, we develop a dual-level representation that incorporates regional consistency into the point-level classifier learning, reducing the noise in generated segmentation. Finally, we conduct extensive experiments on two widely used datasets, SemanticKITTI and SemanticPOSS, and the results show our method significantly outperforms the state-of-the-art by a large margin.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 805